PORTLAND, Ore. — Bubbles are more than a kid's play toy. They may also have powerful uses in medicine and industry. This year's prestigious Gordon Bell Prize has gone to ETH Zurich (Switzerland) and IBM, in collaboration with the Technical University of Munich and Lawrence Livermore National Laboratory, for the world's most complex fluid dynamics simulation of cloud cavitation collapse. The simulation could lead to advances in everything from protecting fluid-handling equipment from damage to destroying cancer cells in humans.
The Association for Computing Machinery (ACM) awarded the prize at Supercomputer 2013 last week in Denver.
"We picked this fluid dynamic problem, because it had a nice mix of applications that were relevant in engineering while at the same time relevant in medicine," Petros Koumoutsakos, director of the Computational Science and Engineering Laboratory at ETH Zurich, who led the project, told us. "Also, because it is very difficult -- almost impossible -- to get the equivalent data from experiments."
The destructive power of collapsing bubbles could destroy cancer cells or shatter kidney stones in human patients.
(Source: ETH Zurich)
Cloud cavitation collapse occurs when bubbles form in a liquid when pressure drops. When these bubble pop, they generate shock waves that can damage equipment, including high-pressure fuel injectors and propellers. These waves also can be harnessed to explode unwanted biological material.
Supercomputers have long been researched as a method of accurately simulating this fluid dynamics problem, but it took massive parallelism to crack the nut. In fact, these researchers set a record for the largest fluid dynamics simulation -- 1.6 million cores and 6.4 million threads. Run on IBM's Sequoia BlueGene/Q supercomputer at Lawrence Livermore National Labs, the simulation achieved 14.4 petaflops of sustained performance -- 73% of Sequoia's peak performance.
The simulation resolved a cloud of 15,000 collapsing bubbles by dividing the geometric space surrounding them into 13 trillion cells. The resolution was 150 times better than the best previous attempt, and the time required was reduced 20-fold.
The researchers will repurpose their algorithm to related fields of science and to relevant applications in engineering and science.
"Now that we have a framework and an algorithmic implementation that allows us to do these simulations efficiently, we would like to extend to engineering problems that are similar to this one, since the area of fluid dynamics is much wider than bubble simulation," Alessandro Curioni, head of mathematical and computational sciences department at IBM Research-Zurich, told us.
IBM Blue Gene/Q Sequoia supercomputer at Lawrence Livermore
National Labs ran the winning bubble simulation.
(Source: Lawrence Livermore National Labs)
Applications in life sciences are also possible, according to the researchers. For instance, advanced therapies are aiming to shatter kidney stones with explosively collapsing bubbles, but they must use trial and error to set the parameters for the procedure. Supercomputer simulations could act as a virtual laboratory for trying different parameter combinations.
"If you have a stone in your kidney, there is a technique that introduces bubbles and then sends in a pressure wave with ultrasound, causing the violent collapse of the bubbles, thereby destroying the kidney stone," Koumoutsakos said. "At the moment, doctors only have qualitative information, but our simulations can give them the quantitative information they need to perfect the technique."
If the technique can be perfected for kidney stones, other research groups could apply it to destroying cancer cells by rupturing them in the presence of cancer-fighting drugs.
The fluid dynamics simulation was chosen to receive the Gordon Bell Award from a field of six finalists narrowed down from hundreds of submitted papers.
— R. Colin Johnson, Advanced Technology Editor, EE Times